first principles of typographic design for document production - tug

TUGboat, Volume 5 (1984), No. 2

First principles of typographic design for document production

Richard Southall

Introduction

Leslie Lamport and I taught a two-day course on 'First principles of typographic design for document production' as a preliminary to the

TUG meeting at Stanford University. August 13-14. 1984.

What follows is an expansion, somewhat revised and restructured, of my lecture notes.

Objectives

I did not feel that i t was possible in two days to teach anything

useful about typographic design on a 'how to' basis (even if the

problems of interpreting conventional typographic design practice in terms of TEX macros had been solved). Nor would i t have

been helpful to give preceptive solutions to a limited set of design

problems. On the old analogy. I did not wish to give away fish to the course members, and I did not have time to teach them how to catch

fish for themselves. In such circumstances, the best thing to do is to try to outline some productive ways of thinking about water and

fishing-tackle, so that basic errors of approach to the problems of

fish-catching can be avoided in the future. Thus I tried to work out, and justify from first principles, a useful way of thinking about the

problems of document design: and to illustrate this way of thinking

by looking at the ways in which typographic designers working in

earlier technologies had tackled similar kinds of problem.

Subject area

The subject matter of my part of the course was written language

Written language has two levels of structure, which may be

called microstructure and macrostructure (Waller, 1980a). The microstructure of written language has to do with details of the

sequence and arrangement of the characters that make up a written

text. It is governed by the rules contained in a 'system of writing':

the set of rules for writing a particular language with a particular

script in a particular technological environment. (I owe this useful concept to John Mountford.) For most systems of writing, and

certainly for those in which the written text is produced by a

mechanical writing-system, these rules are reasonably well defined,

and descriptions of them are readily accessible (Chaundy et al.. 1957;

'Chicago manual of style', 1982: Dowding, 1966; Swanson, 1979:

Walker, 1979): though they differ to a surprising extent between

systems of writing, even in the same technological environment

(Desarmenien, 1984: Walker, 1983).

The macrostructure of written language has to do with its division

into semantic objects (which we can usually recognize, even if we can't make very exact definitions of them - things like

chapters, subsections, paragraphs and list items) and their graphic

embodiment in a document. The rules governing the macrostructure

of written language are much less well defined than those governing


its microstructure. Many typographic designers would deny that

there were any such rules at all (though Twyman. 1981. formulates

'what many typographers would consider one of the few fundamental

"rules" of typography'), because there are so many equally effective

ways of designing a document. But if there are no rules (or almost

none) that designers would agree on. there are some principles for effective design that they (or most of them) would agree with: and

i t was these principles that were the subject area of my part of the

course.

Our discussions in the course were based on two postulates:

Documents have a conceptual structure

Graphic structures can be made that reflect conceptual

structures

one axiom:

The graphic structure of a document should reflect its

conceptual structure

and one desideratum:

The graphic structure of a document should be such that

the document is as easy as possible to use.

The problem of document design for computer-based systems

In the past, the design of documents was done a posteriori.

Books were written before they were designed: the conceptual structure of the author's thoughts was already in place, embodied

in some graphic form or other, and the designer's task was to

render or re-render that embodiment in a semantically effective

and technically practicable way. (See Hewson. 1983. for a detailed

study of the evolution of the graphic embodiment of Wittgenstein's

'Tractatus Logico-Philosophicus' through successive renderings - in

manuscript, typescript and printed form - of the work.)

In the design of documents for computer-based document production

systems, the problem is the other way round. The document has

to be designed a prior;, before the author's thoughts are present

at all. The document designer's task is to devise efFicient graphic

embodiments for conceptual structures that are suitable to f i t any

thoughts that any author using the system might have.

The document designer cannot tell (let alone dictate) what content

an author will put into each of the conceptual structures that the

document design provides for, or in what order the structures will

be used. It is not too hard to provide a graphic embodiment for

each structure, that will behave reasonably if i t is not used in what

its designer would consider to be an unreasonable way; but what

is unreasonable to a document designer may not be at all so to a

mathematician or a philosopher.


2 Typographic structures

Models of text and models of documents

There is a simple 'bottom-up' model of written text, that considers it

as a sequence of characters and spaces.

Sequences of characters make up words words sentences

sentences paragraphs

paragraphs sections sections chapters

chapters books (articles, reports, documents)

This model does not work very well, either as a way of representing

real text or real documents. At the lowest level, i t ignores things like

changes in character style. At a slightly higher level, it doesn't deal at all tidily with quite important questions, like 'What is a sentence?'

(or, indeed, 'What is a word?'). At a higher level still, i t ignores

many of the important things that aren't chapters (in particular, lists

and tables) that go to make up documents.

As a means of understanding the nature of written text, this model

is good for analysing the functions of non-alphabetic characters in

words and straightforward sentences; but otherwise i t is too simple

to be useful with present-day technologies of mechanical writing. I t

was well suited to the train-printer and daisy-wheel era of document

production.

A 'top-down' approach to the modelling of documents begins by looking for recognizable graphic objects, and the semantic objects of

which they are embodiments, in a document.

This approach has been pioneered in the British Library research

project on the graphic translatability of text, with work by

P.E.Norrish in the Department of Typography & Graphic

Communication at the University of Reading. This work has

centred on the study of public information documents, which are

usually much more complex in both graphic and semantic terms than

the technical reports that designers for computer-based systems are

most often concerned with. Norrish and her collaborators have made

very detailed analyses of such documents, looking at the 'access

structures' of headings and paragraph markers that allow users to

find their way around a document, as well as the objects that occur in it.

This work is relevant to the problems of designing computer-

produced documents in two ways. First, i t points up the

considerable complexity, in semantic as well as graphic terms.

of real documents. It is all too easy to think that chapters, sections

and subsections are the only kinds of object that need to be

considered in designing the text of a document. Second, it draws

attention to the large number of alternative graphic forms in which

a particular semantic object can be embodied. Norrish has so far

identified no fewer than forty-seven different types of 'graphic list'

(semantic list structures in which the list items are distinguished by

graphic means).


Section structures

However inadequate the chapter/section/subsection model may be as a way of representing the structure of real documents, i t is very

often used to construct the body matter of technical reports. I f we

look at some of the semantic properties of constructions of this

kind, we can get an idea of the properties their graphic embodiments

ought to have.

Three kinds of relationship exist between the elements of a

chapter/section/subsection construction. There are relationships of

hierarchy: chapters are higher-level elements than sections, which

are higher-level elements than subsections, and so on. There are relationships of containment: higher-level elements contain lower-

level elements. And there are relationships of sequence: objects at

each level follow each other in the construction.

At all the levels of such a construction, the logical structure of the

elements is very similar. Each chapter, section, subsection and so

on contains a heading and a body: the heading relates to the body

that follows i t . The body of the element may contain paragraphs of

text, or lower-level elements, or text followed by lower-level elements,

down to the lowest-level element, whose body (by definition) is all text. Thus there may be 'descending' sequences of headings at the

beginning of an element.

By our axiom, the graphic embodiment of a chapter/section/subsection

construction should reflect its semantic structure. Since the visual

appearance of the paragraphs of text within an element does not

usually alter with the position of the element in the hierarchy (except

perhaps at the very lowest levels), i t is the task of the heading of

each element to show whereabouts in the structure its element is

located.

Thus, the hierarchical relationships between elements should be

clearly expressed in terms of the graphic relationships of their headings; as should their relationships of containment and sequence.

The means that are available to designers for realizing the graphic

expression of such relationships are discussed next.

3 The typographer's tools

Typeface terminology

The terms 'typeface' and 'font', which in earlier technologies had

separate and clearly-defined meanings, are now used more or

less interchangeably in discussions of computer-based document

production systems, to refer to fundamentally different entities.

Great confusion results. The following definitions are proposed, as an aid to clear thinking:

Typeface: a set of distinctive, visually related shapes for some or

all of the characters of a script, intended for mechanical

reproduction

Style: a distinguishing visual characteristic of a typeface


Family of typefaces: a set of visually related typefaces with differing

styles

Font: a set of renderings of some or all of the character shapes of

a typeface, intended for use in a restricted range of output

image sizes in a particular reproducing system

What happens in text?

Written language contains elements which are not alphabetic or numeric characters: punctuation signs, and space. It also has

features which seem to operate at a higher level than the words

of the text: capitalization, changes in type style and size, and the

presence of vertical and horizontal space in varying amounts. These elements and features evidently have some part to play in written

language, and some (but surprisingly few) attempts have been made to identify their functions (Mountford. 1980: Walker, 1979). The

following list of 'roles fulfilled by graphic and spatial features in the

articulation of verbal graphic language' (whose incompleteness is

acknowledged by its author) is taken from Walker's paper:

1 Differentiation

1.1 Emphasis

1.2 Distinction/particularization

1.3 Quotation

1.4 Interpolation

2 Abbreviation

3 Introduction

4 Omission

5 Separation and connection

6 Presentation of numbers

The most significant of these functions seems to be that of differentiation/emphasis.

Graphic conventions and graphic capability

Any mechanical writing-system makes available to its users a certain

graphic capability. This can be expressed in terms of the number of

characters, typefaces and type sizes, and the facilities for defining

amounts of horizontal and vertical space, that the system offers

(Southall. 1982). Until very recently. the graphic capability of the

writing-systems that were available to most computer users was

extremely limited: a single 96-character font of one single-width

typeface, with a single increment of horizontal space the same as the

characters' width, and a single larger increment of vertical space.

Except for the lack of a 'half-line' increment of vertical space, this

capability is roughly the same as that of an ordinary typewriter.

Thus when computer-based document production systems came to

be designed, there were a number of ready-made graphic conventions

for expressing semantic function already available in the rules for the layout of typewritten documents. Almost all of these conventions

could be taken over unchanged into the layout of computer-produced


documents, and there was no need for system designers t o think

explicitly about the semantic functions that were being expressed by

particular graphic configurations.

Laser page-printers, and the first versions of TEX and Metafont.

made writing-systems with enormously increased graphic capability

available t o part of the computing community. The designers of

document production systems did not always understand that

graphic conventions derived from typewriting practice were not

necessarily appropriate for systems with a graphic capability as good

as, or better than, that of conventional typesetting.

The typographer's task

Looking at written language as being made up of semantically

functional graphic elements as well as the words and phrases of

the text, and bearing our axiom in mind, it becomes easy to define

the typographic designer's task. This is to devise efFective graphic

means of expressing the different semantic functions that are

required to embody in a document the conceptual structure of its

author's work.

The way we are accustomed to reading continuous text leads us

to recognize a paragraph as an area of text within which the space

between successive rows of characters does not vary. Because

of this, the graphic means that are available to the designer for

expressing the functions of emphasis and difFerentiation within a

paragraph are limited to those that do not change the space between

rows of characters: capitalization, and changes in type style without

changes in type size.

Outside the limitations of the paragraph, additional graphic means

of expressing semantic function are available. The most powerful

and flexible of these, and the least well understood by untrained

designers, is the use of space Others are changes of type style, and

changes of type size.

4 Making text readable

Legibility research is an old subject (Tinker. 1963) and - as far as

investigations of the typographic requirements for readable text are

concerned - more or less a dead one (Spencer, 1969: Zachrisson.

1965). Current research in reading focusses much more on the

perceptual and mental processes involved (Pirozzolo and Wittrock.

1981: Tzeng and Singer. 1981). Perhaps it is what computer

scientists see as the excessive antiquity of the research on the layout

of readable text that has led it to be so consistently ignored by the

designers of computer-based document production systems. Its

findings, though, are none the less valid for being old, or for being

ignored, and they are as follows:

Lines of text should not be longer than 10-12 words (60-72 characters)

The appearing space between words in a line of text should

be substantially less than the appearing space between

successive lines of the text


The appearing space between words should not vary appreciably from line to line of the text.

These have been the canons of good typographic design for text for

a very long time (Morison. 1951).

5 Designing headings

The functions of headings were discussed in Section 2. Headings

mark the elements they belong to, and they show where those elements belong in the logical structure of the document as a whole.

The latter function almost always means showing two things about

an element: its place in a hierarchy and its place in a sequence.

The graphic hierarchy of the headings in a document should express

the conceptual hierarchy of the elements the headings belong to. Thus the headings of higher-order elements should be graphically

more prominent than the headings of lower-order elements. The

connection between a heading and its element should always be

explicit, so that in a sequence of 'descending' headings it is clear which element a particular heading belongs to.

In principle, the designer has three means for expressing hierarchy

and connection in graphic terms: changes in type size, changes

in type style, and the use of space. Where sequence is expressed explicitly, it is always by some sort of numbering system (Waller,

1980b).

In practice, the availability of graphic means to the designer of a document depends on the graphic capability of the writing-system

the design is being made for. In a system with limited capability,

the graphic means for expressing hierarchy may be exhausted before the bottom of the hierarchy has been reached. In such a case, the

designer may choose to use a more elaborate numbering system

for the headings in the document, so that the numbering expresses hierarchy as well as sequence.

This sort of circumstance, in which the content (in terms of

characters and spaces) of a text is determined by the graphic

capability of the system with which the text is rendered, is an

instance of a major class of problems that have to be faced and

solved by a realistic methodology of generalized document design.

6 Laying out the page

Margins

While our mechanisms of reading have not changed in the last few

years, so that the rules for the layout of readable text can be carried

over unaltered from traditional practice, the same is not necessarily

true of the way we use documents.

The classical canons for the size of margins that are summarized by

Tschichold (1965) apply to reading situations which were usually

very different from the ones in which technical reports are used. Classically, books are codices: they open to a pair of facing pages.

Books intended for continuous reading at normal reading distances


are of such a size that this pair of pages forms a visual unit for the

reader. In these circumstances, a symmetrical layout for the pages

makes sense.

The letter-sized page which is by far the most common in technical

reports is much the same size and format as the quarto page of

traditional book design. Quarto formats were largely used in four

kinds of book: reference books; bibles, in which the main text was

broken up into verses, and the page often crowded with cross-

references and notes; service books, for use in bad light by people who were very familiar with the text: and editions de luxe, where the

width of the margins was often an index of the cost of the book (and

hence the wealth of its owner).

Reference books and quarto bibles are meant for consultation rather

than continuous reading. Service books are intended to be read

at distances that are longer than normal. Editions de luxe are for admiration rather than use. There is no particular justification

for taking over the convention of symmetrically laid-out pairs of pages from such books into the design of technical reports, whose

letter-size pages are seen at normal reading distance (so that the

visual unit is a single page) and are as often as not photocopied and

put in a ring binder (which i t is stretching traditional terminology

rather too far to describe as a codex).

Line length

The use of the letter-sized sheet in laser printers derives from its use in the office, where conventional margins give a line length of six

and a half inches or so: with the character pitch and line spacing

of ordinary typewriters, a line of 65-75 characters and an easily

readable page. The same margins, with the 10 point type that is the

default size in many document preparation systems, give a line of

95-115 characters: far too many to be read effectively, even i f the space between the lines of text is increased (Tinker. 1963). It is true

that making a very long line means that many more characters f i t

on the page, and fewer pages are needed for a document of a given length. The consequent reductions in cost (which, as Leslie Lamport

pointed out during the course, may be more apparent than real) have

to be traded off against the reduced effectiveness of the document as

a means of communication.

In fact, the most efficient way of filling a page with characters is

to use double-column setting. Much narrower margins are tolerable

in double-column than in single-column setting, so that more of

the page area can be covered with type. However, the problems of

designing headings, and of integrating the non-text elements of the

document into the design of the page, tend to become more difFicult.


Other components of the page

Text is not the only thing that appears on the pages of technical

reports. There are footnotes, tables, figures and figure captions,

as well as page numbers and running headlines or footlines. All

these elements need t o be distinguishable in visual terms: figure

captions should not disguise themselves as footnotes, or headlines

as part of the text. The designer needs to pay attention to unusual

circumstances - What happens if a figure falls at the foot of a

page? if a table whose column headings use the same style and size

of type as the headline comes at the head of a page? - rather than laying out the page as if it was only to contain continuous text.

7 Conclusions

There is nothing in the foregoing that is not perfectly obvious as

soon as it is pointed out (except perhaps the requirements for

readable text, and they are in easily accessible research literature).

Equally. I had no difficulty, with little more than a couple of numbers

of TUGboat at hand, in finding many pages whose layout made

them virtually impossible to read, and many instances in which the

conceptual structures of documents were being concealed, rather than revealed, by their graphic embodiment.

Why is this? Why is the average computer-produced document

so hard to use (and such a miserable object, in conventional

graphic-design terms)? The reasons lie in a failure of communication

between computer scientists and graphic designers.

This failure has two main causes. The first is that the designers

of computer-based document production systems most often have

no conceptual apparatus with which to think about graphic design

problems. The thinking of traditionally-educated graphic designers

tends to be unarticulated, visual, and concrete, and their vocabulary oriented towards particular graphic technologies: computer scientists'

thinking tends to be algorithmic, symbolic, and abstract, and

their vocabulary is that of mathematics. Ways of thinking about

document design that computer scientists might find congenial are only beginning to be developed in the graphic design community.

as a response to the problems posed by new information-handling

technologies.

The second cause of failure is at the interface between graphic

designers and computer-based document production systems, and

has to do with the nature of the interface itself. Because graphic designers think visually, they need visual objects to think with; i f the

first task a designer has is to analyse the conceptual structure of a

document, the second is to sketch out possible graphic embodiments for i t. A skilled designer can get a long way by making drawings.

but sooner or later a design has t o be tested by being produced: and i t is at this point, where the specifications on the drawing have to

be translated into instructions to the document production system.

that communication fails. The primitive concepts that are handled by

computer text formatting languages, the terms the languages use t o

describe them, and the commands with which they are manipulated.


are all completely alien to the experience of traditionally-educated

designers.

The result, in the present state of affairs. is that traditionally-

educated graphic designers cannot use computer-based document

production systems. however much they would like to do so. The

most urgent priority. for TEX and systems like it. is for interfaces to

be developed so that designers can bring their skills to bear in an

area where - to say the least - they are badly needed.

References

Chaundy. T.W.. Barrett. P.R. and Batey. C.

The printing of mathematics

London: Oxford University Press. 1957

The Chicago manual of style (13th ed) Chicago: University of Chicago Press. 1982

Desarmenien. J.R.

How to run TEX in a French environment: hyphenation, fonts.

typography (Computer science report STAN-CS-1013) Stanford. California: Department of Computer Science. Stanford

University (1984)

Dowding. G. Finer points in the spacing and arrangement of type

London: Wace, 1966

Hewson, R.L.

The visual presentation of a philosophical text: Wittgenstein's

Tractatus Logico-Philosophicus

Unpublished dissertation: Department of Typography & Graphic

Communication. University of Reading. England (1983)

Morison. S.

First principles of typography Cambridge: Cambridge University Press. 1951

Mountford, J. Writing-system as a concept in linguistics

Information design journal. vol.1 no.4, pp.223-231 (1980)

Pirozzolo. F.J. and Wittrock. M.C. (eds)

Neuropsychological and cognitive processes in reading

New York: Academic Press. 1981

Southall. R.

The problems of forms artwork production

London: Management & Personnel Office. HM Government. 1982

Spencer. H.

The visible word

New York: Hastings House. 1969 (2nd ed)

Swanson. E. Mathematics into type

Providence. R.I.: American Mathematical Society. 1979


Tinker, M.A.

Legibility of print

Ames. lowa: lowa State University Press. 1963

Tschichold, J.

'Non-arbitrary proportions of page and type area'

in Osley. A S . (ed)

Calligraphy and palaeography: essays presented to Alfred Fairbank

London: Faber & Faber, 1965

Twyman. M.L.

Typography without words

Visible language, vo1.15 no.1, pp.5-12 (1981)

Tzeng, O.J.L. and Singer, H. (eds)

Perception of print: reading research in experimental psychology

Hillsdale. N.J.: Lawrence Erlbaum. 1981

Walker. S.F.

The presentation of information about house style: a reconsideration

(IET Text-processing paper 1) Institute of Educational Technology. The Open University, Milton

Keynes. England (1979)

Unpublished thesis: Department of Typography & Graphic Communication. University of Reading, England (1983)

Waller, R.H.W.

'Graphic aspects of complex texts: typography as macro- punctuation'

in Kolers. P.A.. Wrolstad. M.E. and Bouma, H. (eds)

Processing of visible language, vo1.2

New York: Plenum. 1980

'Notes on transforming no.4: numbering systems in text'

in Hartley. J. (ed)

The psychology of written communication: selected readings

New York: Nichols. 1980

Zachrisson, B.

Studies in the legibility of printed text

Stockholm: Almqvist & Wiksell, 1965

Bibliography The publications cited here are chosen, as far as possible, to be

good pointers into the design research literature as well as being

interesting in themselves.

Green. T.R.G. and Payne, S.J.

The woolly jumper: typographic problems of concurrency in

information display

Visible language, vo1.16 no.4, pp.391-403 (1982)

Hartley. J. Designing instructional text

New York: Nichols. 1978


(ed The psychology of written communication: selected readings

New York: Nichols, 1980

Eighty ways of improving instructional text IEEE transactions on professional communication, vol.PC-24 no.1,

pp.17-27 (1981)

Jones. J.C. Design methods: seeds of human futures

London: Wiley-Interscience. 1970

Macdonald-Ross, M. and Smith, E.B.

Graphics in text: a bibliography

Institute of Educational Technology, The Open University. Milton

Keynes, England (1977)

Reynolds, L.

Legibility studies

Journal of documentation. vo1.35 no.4, pp.307-340 (1979)

Spencer, H.

Pioneers of modern typography

London: Lund Humphries. 1969

, Reynolds. L. and Coe. B. Spatial and typographic coding in printed bibliographical materials

Journal of documentation. vo1.31 no.1, pp.59-70 (1975)

Tschichold, J. Asymmetric typography

New York: Reinhold. 1967

Twyman. M.L.

'A schema for the study of graphic language'

in Kolers. P.A.. Wrolstad. M.E. and Bouma, H. (eds) Processing of visible language, vol.1

New York: Plenum. 1979

Waller, R.H.W.

Journal typography in transition

Journal of research communication studies, vo1.3 no.4, pp.335-349

(1982)

Wright. P. 'Usability: the criterion for written information'

in Kolers, P.A.. Wrolstad, M.E. and Bouma, H. (eds) Processing of visible language, vo1.2

New York: Plenum, 1980

"The instructions clearly state . . . . " Can't people read?

Applied ergonomics, vo1.12 no.3, pp.131-141 (1981)